Deep Residual Learning for Image Recognition

Very deep networks used to train worse as layers were added. ResNet has each block learn only a correction that is added to its input through a shortcut, so gradients flow easily and networks with 100+ layers train reliably.

Why it matters

Residual connections are now in almost every deep architecture, Transformers included.

Read the paper: arXiv:1512.03385

Plain-language summary written by SolxLab. Read the full paper at the official link; all rights remain with the authors.